Histopathological Tissue Segmentation of Lung Cancer with Bilinear CNN and Soft Attention

<div>Heatmap of different models generated by Grad-CAM. (a) Slides of lung cancer scanned in 20x magnification factor, (b) heatmap of DeepTissue Net, (c) heatmap of ResNet50, (d) heatmap of the ResNet50 model with bilinear pooling module and attention module. It shows that the ResNet50 model with bilinear pooling module and attention module can detect the largest histopathological region. TUM: tumour epithelium; STR: stroma; LYM: tumour-infiltrating lymphocytes; NEC: necrosis.</div>

BioMed Research International

fig5

Figure 5

Figure 5: Histopathological Tissue Segmentation of Lung Cancer with Bilinear CNN and Soft Attention 